Vehicle Gesture Recognition Using Hand Region Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for recognizing user gestures in vehicles suffer from decreased recognition rates as the number of available hand poses increases, making it difficult to accurately operate vehicle components without diverting the driver's attention from the road.
Innovation Solution
An apparatus and method that detect feature point candidates based on the curvature of a user's hand, using an image generator to create a hand model by filtering candidates, and a gesture recognizer to classify hand poses based on the number of unfolded fingers, improving recognition accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If the number of available hand poses is increased to provide more gesture options, then the versatility of gesture control is improved, but the gesture recognition rate deteriorates
Solution Approach 1:
The hand pose recognition process is segmented into multiple distinct stages: hand region extraction from the image, palm region detection through morphology operations, finger region identification by eliminating the palm region, and feature point detection within each region. This segmentation allows each stage to focus on specific tasks, improving overall recognition accuracy even with increased gesture complexity
Solution Approach 2:
Different detection methods are applied to different regions of the hand based on their specific characteristics. The palm region uses morphology operations (erosion and dilation) optimized for its structure, while finger regions use boundary elimination methods. Feature points are detected with different criteria depending on whether they are fingertips or finger valleys, allowing each local region to be processed with the most appropriate method
Data Source
AI summary
An apparatus for recognizing a user's gesture for carrying out an operation of a vehicle may include: an image generator that generates an object image by photographing a user's hand; a feature point candidate detector that generates a hand image including a hand region by eliminating a background of the object image; a palm detector that detects a palm region by performing a morphology operation on the hand region; a finger detector that detects a finger region by eliminating the palm region from the hand region; a feature point detector that detects the finger valleys located within the palm region and the fingertips located within the finger region as feature points; and a gesture recognizer that recognizes a number of unfolded fingers based on the hand model, and recognizes the user's gesture by selecting a category that corresponds to the number of unfolded fingers among a plurality of categories.